§ HARDWARE · TECHNICAL CATALOG

AI Edge Computing Servers

Comparative technical catalog of certified appliances for Cognitive Server — Dell · Lenovo · HPE · xFusion.

WHAT IS AN AI EDGE SERVER

Computing at the edge with AI acceleration

AI edge computing servers move processing to where data is generated: industrial plants, retail stores, 5G towers, hospitals or remote sites. They cut latency, minimise cloud traffic and preserve data sovereignty. They embed GPU, NPU or dedicated superchip acceleration for on-premise inference.

4 certified vendors

20+ available models

100% on-premise operation

CATALOG BY VENDOR

DELL — PowerEdge XR-Series

The broadest portfolio of short-depth ruggedized edge servers, from the modular XR4000 to the dense XR7620 and the XR8000 RAN platform.

Dell PowerEdge XR4000

Sled-based modular platform with minimum depth

Form factor: 2U, ultra-short depth (~356 mm)
CPU: Intel Xeon D-2700 (up to 20 cores/sled)
Memory: Up to 256 GB RDIMM / 512 GB LRDIMM per sled
GPU: Up to 2× SW 150 W GPU or 1× DW 250 W GPU
Network: 4× 10GbE / 4× 25GbE SFP; OCP 3.0
Management: iDRAC9 + Dell NativeEdge
Use case: Retail, telco, modular far-edge

Dell PowerEdge XR5610

Low-profile 1U rugged for 5G MEC deployments

Form factor: 1U, short depth
CPU: Intel Xeon D-2700 series (up to 20 cores)
Memory: Up to 512 GB DDR5
GPU: Up to 1× single-width PCIe GPU
Network: 2× 25GbE + OCP 3.0; SmartNIC options
Management: iDRAC9 + OpenManage
Use case: 5G MEC, far-edge telco, retail

Dell PowerEdge XR7620

Dual-socket 2U with up to 2 high-performance GPUs

Form factor: 2U, rugged short depth
CPU: 2× Intel Xeon Scalable 4th gen (up to 300 W TDP)
Memory: Up to 4 TB DDR5
GPU: Up to 2× DW 300 W GPU (H100/A100 PCIe)
Network: 4× 25GbE / 100GbE options; OCP 3.0
Management: iDRAC9 with LC; NativeEdge
Use case: Dense inference, machine vision, industry 4.0

Dell PowerEdge XR8000

Modular platform for Cloud RAN and vRAN

Form factor: Modular, ultra-short depth
CPU: Intel Xeon D or ARM Ampere (configuration dependent)
Acceleration: Intel/Xilinx FPGAs for L1 vRAN
Network: High-density fronthaul interfaces
Management: iDRAC9 + NativeEdge + OpenRAN compatible
Use case: 5G vRAN, Cloud RAN, telco far-edge

Dell PowerEdge XR9700

The most powerful edge server in the XR range

Form factor: 2U rugged
CPU: Up to 2× Intel Xeon Scalable 5th gen (Emerald Rapids)
Memory: Up to 8 TB DDR5
GPU: Up to 4× high-performance PCIe GPU
Network: Up to 400GbE; OCP 3.0
Management: iDRAC9 + NativeEdge
Use case: Edge HPC, on-premise generative AI, aggregation

CROSS-VENDOR COMPARISON

All 20 models in one view

Vendor	Model	Form factor	CPU	Max memory	Max GPU	Primary use case
Dell	XR4000	2U modular	Xeon D-2700	512 GB	2× GPU SW 150W	Modular far-edge
Dell	XR5610	1U rugged	Xeon D-2700	512 GB	1× GPU SW	5G MEC, telco
Dell	XR7620	2U rugged	2× Xeon Scalable	4 TB	2× GPU DW 300W	Dense AI, industry
Dell	XR8000	Modular	Xeon D / ARM	—	FPGA L1	vRAN, Cloud RAN
Dell	XR9700	2U rugged	2× Xeon 5th gen	8 TB	4× GPU PCIe	Edge HPC, gen AI
Lenovo	SE100	Mini/desktop	Core Ultra NPU	64 GB	Integrated NPU	Retail, IoT
Lenovo	SE350 V2	1U rugged	Xeon D-2700	256 GB	1× GPU SW	Far-edge, manufacturing
Lenovo	SE360 V2	2U silent	Xeon D-2700	256 GB	1× GPU SW	Office, retail
Lenovo	SE450	1U	Xeon Scalable	1 TB	3× GPU SW (L4)	AI inference
Lenovo	SE455 V3	1U	AMD EPYC Siena	3 TB	6× GPU SW / 2× DW	On-premise LLM
HPE	DL145 Gen11	1U	AMD EPYC Genoa	1.5 TB	3× GPU SW 75W	Near-edge, NFV
HPE	EL8000	4U modular	Xeon Scalable	variable	GPU + FPGA	Industrial, 5G MEC
HPE	e930t	1U rugged	Xeon D	256 GB	1× GPU SW	Telco far-edge, vRAN
HPE	e920d	Desktop	Xeon D	128 GB	—	Retail, branches
xFusion	GB10	Mini-PC	NVIDIA GB10 SoC	128 GB	1 PFLOP FP4	On-device AI, local LLM
xFusion	1288 V8	1U	Intel Xeon 6	2 TB	3× GPU SW	Near-edge, NFV
xFusion	2288 V8	2U	2× Intel Xeon 6	8 TB	6× GPU SW / 4× DW	Dense AI, HPC
xFusion	G5500 V7	4U	2× Xeon Scalable	4 TB	8× GPU DW	Near-edge LLM training
xFusion	G5200 V7	2U	2× Xeon Scalable	2 TB	4× GPU SW	Inference, NLP
xFusion	XtationX3	Tower	Xeon W / TR PRO	512 GB	2× GPU pro	Creative edge

SELECTION GUIDE

The right appliance for each deployment

UC1

Low-power AI inference (store / office)

Local processing without datacenter infrastructure. <100W consumption. Optional fanless.

ThinkEdge SE100
PowerEdge XR5610
HPE e920d
FusionXpark GB10

UC2

Dense AI acceleration (industrial plant / aggregation)

Intensive GPU workloads. Extended temperature. High availability.

ThinkEdge SE455 V3
PowerEdge XR7620
FusionServer 2288 V8
HPE EL8000

UC3

Telco, 5G vRAN and far-edge

Short profile, ruggedized, high-density fronthaul. OpenRAN compatible.

PowerEdge XR8000
PowerEdge XR5610
HPE e930t
ThinkEdge SE350 V2

UC4

On-device AI agent development and deployment

Local LLM, light fine-tuning, rapid prototyping. No cloud dependency.

FusionXpark GB10
ThinkEdge SE455 V3
PowerEdge XR9700

UC5

Near-edge HPC and on-premise generative AI

High GPU density, InfiniBand/400GbE, inference / fine-tuning clusters.

FusionServer G5500 V7
PowerEdge XR9700
ThinkEdge SE455 V3

Need help picking the right appliance?

Our engineering team helps you select, size and certify the right hardware for your Cognitive Server deployment.

Talk to engineering →